SongSmith is a prototype web app designed to help novice singers practice solo vocal performance through accessible, AI-assisted feedback. Built with Python, React.js, and Flask, the tool provides visual and text-based feedback on pitch and volume by analyzing user recordings and aligning them to reference melodies.
The initial version included basic pitch tracking, volume visualization, and LLM-generated vocal coaching. While early in development, SongSmith demonstrated strong potential for expansion—particularly in areas like rhythm tracking, group feedback, and guided practice based on deliberate learning principles.
This project laid the groundwork for a more comprehensive, scalable system and deepened my experience with audio processing, user-centered design, and the integration of generative AI in creative tools.